Research On Short Texts Based Internet Users' Intention Recognition And Application

The rapid development of Web 2.0 has greatly improved the communication speed and the scope of communication of Internet users, but also provides a convenient way for Internet users to participate in discussion and express a view on events. In daily life, people not only use social media to accept or publish the creation of a variety of information, but also to express themselves on the sale of goods, recruitment and employment, rental housing and wanting to rent a house intent, which is a kind of service demand intention. People can express their views towards social events through social media, and in order to express themselves to participate in an event participation intention. According to the China Internet Network Information Center(CNNIC) thirty-seventh "development Chinese Internet Statistics Report" released January 22, 2016 survey pointed out that social networking applications are divided into two categories: one is the vertical social networking application more professional niche segments, such as interactive information websites and forums; the other is a comprehensive social information relates to various applications, such as micro-blog. The two major categories of social applications mentioned in the report are the main concerns of this paper. How to use these huge amounts of data generated by Internet users in the social application is an urgent problem to be solved. In this study, we aimed to identify user intent, including the intention of the vertical social networking application and the intention of the comprehensive social information relates to various applications.First of all, for the service of vertical social networking application user demand intention recognition, we design intention recognition model named Verb Biterm Topic Model(Verb-BTM). This model takes the short texts of natural language of users as the research object, with the model, the short texts publised by users will be expressed as template in the form of intent. The formation of intent template is divided into two steps: the first step, take short text without the verb as the intention of the object, using Biterm Topic Model(BTM) for topic mining; the second step take the verb as a template with Vecter Word(word2vec) clustering analysis. Then, after the intention recognition of service demand, in order to meet the needs of the users and improve the user experience, we designed corresponding intention matching method. The method is based on the intention of pair wise definition, and uses dynamic word expansion technology to expand the intention words in order to achieve the objective of the optimization results.Secondly, on the comprehensive social application of the event participants intention recognition. We take the microblog data as the research object, and form multi feature analysis model with the characteristics of the content of the text, text emotional features, text social attitude to discover the tendency expression through participating in the event. The effectiveness of the method is verified by the difference test, the comparison of the results of the public opinion analysis with the people and the regression experiment.Through the above two steps, we prove that the use of short texts for Internet users' intention recognition in two kinds of social applications can receive a certain research results. Finally, in order to visualize the two types of intention recognition results, this paper designs an intention recognition recognition system, which is based on the theory of our research, combined with the two types of application scenarios.